Soil erosion is caused by different geomorphic process such as running water, wind and glaciers. Several methods and equipments have been developed up till now to measure soil erosion caused by these processes but none of them is free of demerits and 100% accurate. However 'soil erosion test-plots method is commonly used. For this kind of measurement test plots should be equipped with runoff measurement and sampling systems. For flow measurement, H-flume, long throat flume, drop box, and V-notch weir are usually used.
Another device namely 'pin-meter' has also been developed to measure soil erosion and deposition. Furthermore, 'devices and instruments for measuring sediments' by FAO can also be followed at website
There are different methods for soil erosion measurement at the watershed scale. The best one depends on some factors such as the availability of the specific types of equipment, or some governing characteristic on your study watershed.
Firstly, using the data of sediment yield in the outlet of the watershed could be a reliable indicator of soil erosion.
Using the data recorded for the erosion pins if available is another option.
Experiment plots could be very useful tools in this regard, too. I suggest this option if it is possible for you.
Additionally, using some physical models such as RUSLE to predict soil erosion but not for measuring.
Finally, there are some innovative methods such as 137Cs.
Runoff plots are better to measure erosion specific to a given land use type and form of erosion. The outlets of a watershed (Mini-to big watershed) is difficult to determine the response of a specific land use type and management conditions. I want to compare these two scenarios in my future research proposal.
There are many different man-induced erosion processes. These processes operate at different scales, spatially and temporally. Measurements must be proper to the scale.
Slot samplers can be used for any erosion studies from small test plots to actual watersheds of several km2. These samplers are automatic and no observer is required during the sampling operation.